On Tue, Jul 23, 2019 at 7:49 AM Giulio Fidente gfidente@redhat.com wrote:
On 7/11/19 7:52 PM, Emilien Macchi wrote:
Even though TripleO is well known for its maturity, it also has a reputation of being complex when it comes to the number of tools that it uses. Somewhat related to the efforts that James is leading with "Scaling TripleO" [1] [2], I would like to formalize our joint efforts to make TripleO simpler in the future. Some work has been done over the last releases already and yet we have seen net benefits; however we still have challenges ahead of us.
- With no UI anymore, do we really need an API?
- How can we reduce the number of languages in TripleO? ... and make
Python + Ansible the main ones.
- How can we reduce our dependencies?
I created a document which explains the problems and propose some
solutions:
https://docs.google.com/document/d/1vY9rsccgp7NHFXpLtCFTHQm14F15Tij7lhn5X_P1...
For those who can't or don't want Google Doc, I've put together the notes into etherpad [3] and I'll take care of making sure it's updated at last at the beginning until we sort things out.
thanks Emilien,
I added a couple of comments in the doc related to two topics which I think have been half-baked in past releases and could be reviewed with the removal of mistral, specifically:
- TripleO invocation environment files should not be order dependent
- Performing a stack update should not require all original environment
files
not sure if people familiar with workflows would be interested in helping with these two?
I'm happy to help. Maybe we can sync up in IRC this week to discuss the environment ordering. That was a feature we added last year after some discussion.
-- Giulio Fidente GPG KEY: 08D733BA